A new server's first week, as its owner sees it: the player graph sits on zero most of the day, climbs to two or three when friends log in after work, and drops back overnight. That is the ordinary shape. Of the 9,319 servers JoinGG measured in one week, 47.5% averaged under one player. The top 1% held 44.0% of all player-hours, and the whole bottom half held 0.7%.
How the week was measured
Every 15 minutes JoinGG's poller asks each listed server how many players it has, in whatever protocol the game itself uses: the Server List Ping for Minecraft, A2S_INFO for Valve's games, the published HTTP endpoints for FiveM. An hourly sample of those answers is stored. The window here is the 168 hours from 10:00 UTC on 21 September to 10:00 UTC on 28 September 2026, and each server's figure is its average player count across those readings.
Only servers that answered at least once are included, and Minecraft listings on free hosts that answer the ping while no server is running behind them were left out. "Player-hours" is the sum of the hourly counts, so one player online for an hour is one player-hour. For the Source and GoldSrc games (the Counter-Strikes, TF2, Garry's Mod, Left 4 Dead 2, Half-Life), the count includes bots, because A2S reports them inside the player total. The methodology page has the details. Every number is about the servers JoinGG lists, not every server a game has.
How uneven it is
Across all ten games together, the top 1% of servers held 44.0% of player-hours. The top 10% held 80.8% and the top 20% held 90.7%. Everything in the bottom half of the list, taken together, came to 0.7%. The median server averaged 1.22 players.
An average under one player does not mean a server was always empty. A server with four people on for three evening hours a day, and nobody otherwise, averages 0.5. What the figures do show is where the week's players were: on a small group of servers at the top.
Game by game
| Game | Servers | Top 1% hold | Top 10% hold | 10 biggest hold | Averaged under 1 player | Never had a player | Median server average |
|---|---|---|---|---|---|---|---|
| Minecraft | 3,118 | 60.4% | 93.1% | 37.3% | 64.4% | 28.5% | 0.21 |
| FiveM | 1,598 | 19.0% | 63.4% | 13.7% | 19.3% | 2.1% | 9.30 |
| CS 1.6 | 1,470 | 10.2% | 41.4% | 8.7% | 31.2% | 10.3% | 3.02 |
| CS2 | 836 | 7.6% | 41.4% | 9.0% | 40.6% | 7.9% | 1.60 |
| CS:S | 553 | 6.6% | 40.0% | 10.2% | 39.1% | 15.6% | 3.92 |
| TF2 | 550 | 12.3% | 50.9% | 17.1% | 58.2% | 12.9% | 0.44 |
| Left 4 Dead 2 | 437 | 13.7% | 63.2% | 26.8% | 75.3% | 13.7% | 0.24 |
| Garry's Mod | 327 | 11.2% | 67.4% | 30.7% | 61.5% | 20.2% | 0.16 |
| CS:GO (legacy) | 233 | 10.0% | 50.1% | 32.0% | 52.4% | 16.3% | 0.97 |
| Half-Life | 197 | 6.0% | 48.9% | 28.2% | 64.0% | 25.9% | 0.23 |
"Hold" is the share of that game's player-hours. "Never had a player" means zero in every reading in which the server answered.
Minecraft is a few big names and a very long tail. Its top 1% held 60.4% of the game's player-hours and its top 10% held 93.1%, while the median server averaged 0.21 players and 28.5% of servers never showed a single player all week. Most of the servers at the top are networks: several servers behind one proxy address. The Minecraft stats page shows the live numbers.
FiveM has the opposite shape. Its top 1% held 19.0%, the median server averaged 9.30 players, only 19.3% of servers averaged under one, and just 2.1% never had anyone on. Most FiveM servers also have gaps in their hourly readings (only 15 of the 1,854 had a reading at every hour of the day that week), so FiveM's averages rest on fewer readings than the other games'.
CS 1.6 and CS:S have the biggest middle. Their top 1% held only 10.2% and 6.6%, their top 10% about 40%, and their median servers averaged 3.02 and 3.92 players, the highest after FiveM. Some of that middle may be bots: a server that keeps bots running around the clock shows a steady few players whether or not anyone is playing. What a number on a server list can and cannot tell you covers the other ways a count misleads.
CS2 sits below the two older Counter-Strikes, with a median of 1.60 and 40.6% of servers under one player. TF2, Left 4 Dead 2, Garry's Mod, the legacy CS:GO servers and Half-Life all have medians under one, and in each of them the ten biggest servers alone hold between 17% and 32% of the game's player-hours. Left 4 Dead 2 has the emptiest middle of all, with 75.3% of servers averaging under one player, which fits a co-op game where a server can sit idle until a group takes it.
If you run a server
The concentration is the reason a copy of a big server rarely works. People who want a large Minecraft network are already on one of the few that hold most of the players, and the ten biggest Minecraft servers alone held 37.3% of the game's player-hours. A new server with the same format asks those players to leave something that works. A niche asks a smaller group for something they cannot find yet: a mode few servers run, a language, a time zone, a rule set nobody else keeps. Most of the reasons nobody joins a new server are versions of this problem.
The other half is that an empty server tends to stay empty. Players open a list, see zero, and join a server that already has people on it, and each of those choices adds to the servers that were busy anyway. Seeding a new server covers getting past zero: a small group playing at fixed times, so that anyone who looks in during those hours finds a game. After that, turning first visitors into regulars is slower work.
If you are looking for a game
The top 10% of servers held 80.8% of player-hours, so sorting by players finds a busy game quickly. What that sort hides is the middle: servers with a handful of regulars that never reach the top of any list, where a new name gets noticed. A quiet afternoon on one of those is not the same thing as a server in decline, and there is a separate guide on telling a dying server from a quiet one.
On the Source games, remember the bots. Six players on a CS 1.6 server at four in the morning may be six bots.
Lists that rank by players
Any list sorted by player count mostly reflects the concentration above. The same few servers hold the top places in every game because they hold most of the players, and a server averaging 0.21 players, the Minecraft median, will not appear near the top however well it is run. A player count is one fact about a server, best read next to its uptime and how long it has been listed.
For a server that opened this week
Measure it against the median for its own game rather than the top of the list. A new Minecraft server averaging one player across its first week is already above the 0.21 median; a FiveM server would need to average 9.30 to reach the middle of its game. Expect most hours to show zero at first. Put your seeding into one or two fixed evening windows instead of spreading it thin, keep those times for a few weeks, and judge the server on its first month as a listing. A first week says very little.
- Tags
- player countsserver listsserver data
- Share
Published · 6 min read